Who Are You Actually Working For When You Write a Newsletter?

Most people don’t get stuck on ideas. They get stuck because every piece starts from zero. Open trending topics, copy three headlines, dig for background, feed it to ChatGPT for an outline, rewrite the tone yourself, make a cover in Canva, paste into your email platform, adjust spacing, hit save draft. Two to three hours gone. Worse: the intro you rewrote, the parallel structure you hate, the evidence boundaries you insist on—all reset to zero next time. The AI acts like it just met you.

Running a daily newsletter with this manual grind is paying a tax on repetitive labor. If you want ad revenue and affiliate income, you need at least 20 publishable pieces a month. At 2.5 hours each, that’s 50 hours. Plenty of people quit by month three—not because traffic died, but because the process buried them. Updates stopped, sponsors walked.

The bar for newsletter monetization was never “can you write.” It’s “can you write consistently, sound like yourself, and actually hit publish.” Miss any one of those and you’re spinning your wheels.

What Is WeWrite: A Tool That Writes Your Newsletter for You

WeWrite is a tool that automates newsletter writing. You tell it “write a piece about X” and it handles sourcing, drafting, formatting, and saving to drafts. It’s not built for a regular web editor—it’s built for AI agents.

For non-technical readers: An AI agent is an AI assistant that can operate your computer. You don’t need to code. You just talk to it. If you have zero technical background, the easiest entry point is telling Claude: “Install this tool for me: https://github.com/imraywang/wewrite.” It’ll run the whole setup.

This tool has over 3,000 downloads on GitHub. Not a toy—a maintained, mature project. The repo is at https://github.com/imraywang/wewrite. Per the README: 3,286 stars, 521 forks, last push 2026-09-07. The official description is blunt: a full-pipeline newsletter tool, from trend scraping to draft folder, one sentence runs the entire content flow.

The workflow is fixed: scrape trends → score topics → article brief → claims and source list → safety content enhancement → first draft (with real research + style injection during writing) → editorial review → revise and re-review if needed → final delivery.

This design is nothing like the “one mega-prompt to rule them all” approach. Images, formatting, and draft pushing are all optional and all downstream. Formatting and publishing never secretly trigger image generation. The original text is always preserved; images generate as separate files plus an illustrated copy. You edit your way, then tell it to “learn my edits,” and the next piece sounds more like you.

Every stage is an independent tool. Every article saves to a local folder. Progress is resumable. Multiple pieces run in parallel without overwriting each other. Pick topics in the morning, say “continue where I left off” in the afternoon. For a daily newsletter, resumability beats “perfect one-shot generation” by a hundred times—you leave for a client lunch, come back at night, and don’t have to rerun the whole chain.

The source post also claims it does “SEO based on real search data.” I haven’t independently verified that. What the README confirms: the writing stage searches real sources, drafts from real material, and explicitly forbids fabricating author experiences. If review fails, it revises and re-reviews. For anyone relying on both search and subscription traffic, this matters more than any “SEO miracle” slogan—readers want verifiable judgment, not fluent filler.

How to Use It: Three Steps

Step 1: Install (one command, copy-paste, done)

WeWrite offers three install paths. Pick whichever fits your environment. The easiest: tell any AI agent, “Please install https://github.com/imraywang/wewrite (run the install.sh in the repo).”

If you want to do it yourself, the recommended path is the one-liner script: shallow-clone the repo to your home directory, run install.sh. The script does three things: installs the CLI tool, symlinks 10 tool modules into Claude Code and OpenClaw directories, and migrates old user state to the new location.

1
2

git clone --depth 1 https://github.com/imraywang/wewrite.git ~/wewrite
cd ~/wewrite && bash install.sh

Want just specific modules without cloning the whole repo? Use: npx skills add imraywang/wewrite. Modules are self-contained, copy-and-use. The CLI installs separately: uv tool install wewrite (or pipx install wewrite).

Don’t rush to configure your email platform after installing. Writing requires zero API keys. Formatting generates local HTML. If you request images but have no image service, it outputs image prompts you can run through Midjourney or another generator. Only when you actually want to push to drafts do you need to copy the config to ~/.wewrite/config.yaml and fill in your platform credentials. The cost switch stays in your hand—far more respectable than locking you into an expensive model from day one.

Step 2: Start With “What Should I Write Today?”

The practical way to use this is to break the pipeline into separate commands, not “produce one complete piece” every time. Tell the agent “what should I write today” or “find some topics” and it outputs 10 scored, ranked topics. Say “write a piece on this topic” and it produces a brief, claims and source list, then a first draft. Say “check this” and it runs fact-checking plus necessary revisions, then delivers the final piece with an editorial report. For covers, say “make a cover for this piece”—it generates one cover without touching the original text. For full illustrations, the same module adds inline images and an illustrated copy. Push to drafts or switch themes with “push to drafts”—if conditions aren’t met, it keeps the local HTML and never publishes without your say-so.

This “one step at a time” approach is critical for monetized newsletters. Trend pieces need speed: confirm the topic, write, review, push—20 minutes to draft, hitting your scheduled send window. Evergreen pieces need evidence: lock down the source list at the brief stage. If a data point doesn’t check out, swap the claim right there instead of rewriting 3,000 words later. Review-style pieces need personality: 7 writing personas switch like themes, from late-night friend to cold analyst, one config line changes the voice. Interactive mode pauses at topic or outline stage for confirmation, preventing the agent from picking a controversial topic you never wanted to touch.

Formatting has 18 themes, all inline styles, WeChat-compatibility fixes, and dark mode. After installing, run wewrite gallery—browser opens with side-by-side comparison, one-click copy. The repo docs render the same sample article in 6 long-form layouts, including step cards, callout boxes, timelines, quotes, summary components, and AIGC disclosure footnotes. learn-theme can learn a new theme from any newsletter. Find a competitor whose layout you like, feed it one piece, and your next piece formats in that style. The most painful part of email platforms is format breakage—this engine turns “hand-tune CSS” into “pick a theme and preview.”

Step 3: Edit Once, and It Writes Your Preferences Into Rules

WeWrite’s most valuable feature isn’t “it can write.” It’s “it gets more like you with every use.” Three mechanisms stack: it remembers your edit patterns, it builds a style library from your samples, and it tracks which pieces get read most and writes more of those. After you edit a piece, tell the agent “learn my edits” and it converts your changes into rules. Import samples through the same module—the style library remembers your preferred sentence length, judgment framing, and banned words. It also feeds reading data back in: which topics get high open rates, which get low completion rates, and the next round of topic scoring adjusts accordingly.

The math is straightforward. Week one, you might still spend 40 minutes editing each piece because the rule library is empty and the agent only knows generic “newsletter voice.” Week two, after feeding it 5 of your finalized pieces, it starts avoiding the empty jargon you hate and opens with judgment-first framing. Week three, trend pieces only need you to adjust data and examples—the tone already sounds like you texting a friend at night. At that stage, per-piece human time drops to 15–25 minutes: 3 minutes confirming the topic, 8 minutes scanning facts, 5 minutes on cover and theme, 2 minutes pushing to drafts. Compared to 2.5 hours of manual work, that’s roughly 2 hours saved per piece. At 20 pieces a month, 40 hours back. Spend those 40 hours pitching sponsors, building your email list, selling a template pack on Gumroad—that’s the actual business structure a newsletter should have.

Safety content enhancement and quality self-checks sit at the writing and review stages. It requires clarifying the reader’s problem, core judgment, and evidence boundaries before drafting; fabricating author experiences is forbidden. This matters enormously for sponsored content—brands fear pieces claiming reviews you never did or factories you never visited. The review report lays out every failing point. You’re editing judgment, not polishing adjectives. Long-term, this beats any “remove AI traces” filter: traces come from emptiness, not punctuation.

How to Make Money With It: Multi-Platform Publishing, Client Work, Paid Columns

Publishing the newsletter is only the first half. Take the same final piece and tell the agent “rewrite for Lemon8” or “rewrite for TikTok.” It does content-level rewriting, checks editorial quality and source similarity—not just chopping long text into short sentences. Lemon8 needs cover hooks, segmented paragraphs, and comment-section copy. TikTok needs spoken-word rhythm, a 3-second hook, and visual cues. Your newsletter piece is “judgment + evidence.” The Lemon8 version becomes “scenario + conclusion + action.” The TikTok version becomes a script you can read aloud. Source similarity stays low, so platform moderation and readers won’t instantly flag it as recycled.

Practically, I recommend a fixed rhythm: publish the deep-dive newsletter at noon, produce the Lemon8 visual post in the afternoon, and the TikTok script in the evening. Three platforms carry the same judgment, capturing both search and recommendation traffic. If a competitor account is already big on Lemon8, use learn-theme to learn their layout and “learn my edits” to learn their sentence density—faster than reverse-engineering “viral formulas” yourself by an order of magnitude. Affiliate-driven accounts can use the newsletter as the trust layer, Lemon8 as the product entry point, and TikTok as ad creative source. Sell the same judgment three times, amortize content costs—that’s real multi-platform publishing, not copy-paste with a new cover.

Digital products work the same way. Use WeWrite to consistently produce a column. After 20 pieces, package them into a paid column, template pack, or prompt library. List it on Gumroad for $29 or $99, link it at the bottom of your newsletter. Your readers have already been educated by your judgment—conversion beats cold traffic by a tier. Freelancers can use this as a delivery tool: client gives you an industry topic, you deliver a newsletter piece + Lemon8 post + TikTok script in 40 minutes. Your quote goes from $300 per piece to $800 per package. Tool cost is API fees. You’re selling process and editorial judgment.

Action List: Get Your First Piece Out Tonight

The following is an estimation model, not actual data. Assume you run a career/side-hustle newsletter targeting daily publishing, sponsorships, and paid columns. Manual era: 30 minutes topic selection, 40 minutes research, 60 minutes writing, 20 minutes tone adjustment, 30 minutes cover and formatting—about 180 minutes total, often with format rework in the email platform. With WeWrite: 10 minutes topic confirmation, 15 minutes writing + review to final draft, 10 minutes checking evidence yourself, 8 minutes cover and theme preview, 2 minutes pushing to drafts after approval. About 45 minutes total. At 22 pieces a month, that’s 66 hours down to 16.5 hours—roughly 50 hours saved. Previously 2.5 hours per piece, now 45 minutes, saving 40+ hours a month.

Three steps tonight:

1. Run the install script (or tell your agent “install this tool for me”)
2. Tell the agent “what should I write today”
3. After editing your first piece, immediately say “learn my edits”

Every day your rule library sits empty, you pay another day of labor tax. The tool is open source. The personality has to come from you. Go to https://github.com/imraywang/wewrite, install it, and get that first piece into your drafts tonight. That’s when your newsletter starts compounding.